Lawrence Livermore National Laboratory's (LLNL) Livermore Computing Complex will be the home of Sierra, the lab's newest high-performance supercomputer, which is scheduled for acceptance in fiscal 2018.
The 125-petaflop/second system is expected to top the sustained performance of Sequoia, LLNL's current main supercomputer, by a factor of four to six.
Sierra came from the U.S. Department of Energy's Collaboration of Oak Ridge, Argonne, and Livermore (CORAL) partnership, culminating in the delivery of large-scale, high-performance supercomputers at each of the labs.
LLNL chose IBM to deliver Sierra to meet the mandates of the National Nuclear Security Administration's Advanced Simulation and Computing program.
Sierra will have two IBM Power 9 processors and four NVIDIA Volta graphics processing units (GPUs) for each node, with the Power 9s supplying memory bandwidth from the chips to Sierra's DDR4 main memory.
LLNL's workload stands to benefit from second-generation NVLINK, supporting a high-speed connection between the central processing units and GPUs, says LLNL's Bronis de Supinski, recipient of the ACM Gordon Bell Prize for 2006.
From Lawrence Livermore National Laboratory
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA